Energy 111 (2016) 818e829

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Diffusion of renewable energy technologies: The need for policy in

Colombia
Maritza Jimenez a, Carlos J. Franco a, Isaac Dyner a, b, *
a
b

Universidad Nacional de Colombia, Colombia

Universidad Jorge Tadeo Lozano, Colombia

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 22 January 2016
Received in revised form
1 June 2016
Accepted 8 June 2016

Although the diffusion of solar electricity has been swift worldwide, there is little certainty regarding the
adoption of this technology on rooftops, in locations where there is no specic policy on renewables in
place. In such cases, there is much uncertainty regarding the effect of solar penetration on electricity
prices to consumers and on the reduction of electricity demand from the grid. The penetration of solar in
the residential sector e which in some cases accounts for about 40% of the country's total electricity
demand e may have a tremendous impact on incumbent utilities and the industry as a whole. Much
research has been devoted to assessing the effect of policy on the diffusion of renewables but not much
work about the developing world is known, particularly in those nations where institutional arrangements do not favour these technologies. Using system dynamics, this paper examines these issues,
considering the diffusion of rooftop solar both with a battery support system, and also without any type
of storage system. An important conclusion is that policy is essential for system sustainability when PV
diffusion is taking place.
2016 Elsevier Ltd. All rights reserved.

Keywords:
Photovoltaics
Renewables
Technology diffusion
System dynamics

1. Introduction
An international outlook on the progress of Photovoltaics
(hereinafter PV, see Table A1 for all abbreviations), in recent years,
shows that this alternative for energy generation has been
increasing in both cumulative capacity (GW) and in generation
capacity (TWh). The International Energy Agency (IEA) [1], in its
Solar Photovoltaic Energy report, shows that the cumulative global
PV capacity has been increasing at an average rate of 49% per year,
in recent years. Moreover, the IEA's Medium-Term Renewable Energy
Market Report 2014 [2] shows how PV generation is increasing in
different regions and that this trend is expected to continue in the
near future.
Rapid increases in cumulative and generation PV-capacity can
be explained three ways: rst, by the intrinsic benets of PV
technology, which have motivated its adoption because of its
simple power generation process [3] and its environmental
friendliness [3,4]; second, by the reduction of PV systems' production costs [4], as the technology learning curves have made

* Corresponding author. Carrera 80 No. 65-223, Bloque M8A, Medelln, Colombia.

E-mail addresses: mjimenezz@unal.edu.co (M. Jimenez), cjfranco@unal.edu.co
(C.J. Franco), idyner@unal.edu.co (I. Dyner).
http://dx.doi.org/10.1016/j.energy.2016.06.051
0360-5442/ 2016 Elsevier Ltd. All rights reserved.

impressive progress since the beginning of the century; these

reducing costs are expected to continue to progress in the midterm [5], which will further improve the perceived benet of PV
systems; and lastly, as policy is seeking alternative power generation technologies as a way to face different issues regarding environmental problems, geopolitical instability and price volatility of
fossil fuels [6], policy has played a signicant motivating role for
rooftop adopters, by subsidizing the implementation of PV systems
[3].
Despite the benets of solar generation, governments are concerned about the rapid increases in the technology's diffusion for
two reasons: rst, as this has exceeded initial expectations in a
number of countries that implemented favourable policies, such as
the United Kingdom, Spain and Germany [7e11]; a second concern
is that this growth has affected the stability of the interconnected
system, prompting redenitions of the incentive structures in order
to control solar diffusion and its effects on market stability
[2,12e14]. However, in many regions in the developing world,
where poor people have little access to electricity, there is
considerable opportunity to take advantage of this technology.
Against the preceding background, this paper examines what
may happen in electricity markets where policy is being considered
or is in early stages of implementation, by exploring unintended

M. Jimenez et al. / Energy 111 (2016) 818e829

and undesirable consequences by means of simulation experiments. To examine the issue, this paper develops a simulation
model built to assess, under two scenarios, the effects of the
diffusion of PV in the residential sector of an electricity market,
considering: rst, the standard diffusion of PV systems that maintain some support from the electricity grid and, second, PV systems
with storage support but no grid support, which provides the
householder with grid independence.
Colombia has been selected as the application case, given that
this is a country with favourable conditions for PV generation e PV
is cost competitive with the grid tariff [15] e and also that despite
the recently promulgated Law 1715 [16], which favours renewables,
there is much uncertainty concerning its implementation and
particularly regarding its effect on the rate of adoption, on prices
and on security of supply issues in the Colombian energy market.
The focus is on the residential rooftop solar generation sector in
Colombia as: a) the residential sector contributes about 40% of the
total electricity demand [17], b) 70% of all dwellings are houses [18],
and c) a large percentage of dwellings can easily install microgeneration PV systems. Some of the lessons from the Colombian
case might also be applicable in other situations and countries.
To reach the proposed objective, this paper is organized as follows: Section 2 provides a review of the international experience
on renewables' diffusion, Section 3 describes the Colombian energy
market and Section 4 describes the implemented system dynamics
(SD)-based model. Section 5 presents the main simulation results,
Section 6 reports validation and conducts a sensitivity analysis, and
nally Section 7 presents the conclusions of this work.
2. International experiences of the diffusion of renewables
A number of papers examine the impact of the diffusion of renewables on electricity markets, from aspects such as: consumption habits, supply intermittency and transmission system balance.
In this direction, Balcombe, Rigby & Azapagic [19] indicate that the
high diffusion of renewables in the United Kingdom has caused
problems regarding grid intermittency and market balance; similarly, der Veen & De Vries [20] illustrate market efciency and
reliability problems in the Dutch case. In the Spanish context, while
utilities experience reductions in revenues and prots, households
n [22] and
experience increases in electricity tariffs [21]. Wide
Keirstead [23] show that while in Sweden users change their consumption habits by increasing electricity consumption when using
renewables, conversely, United Kingdom (UK) households using
renewables reduce their electricity demand.
Many countries use different policies to promote renewables
[2,24]. Feed-in Tariffs (FITs) have been used with good results in
Japan, Germany, the United Kingdom and Spain, among many
others [3,7,9,10,25,26]. Other incentives, elsewhere, have focused
on nancial incentives such as Green Certicates [7,10] or Net
Metering [21].
Nevertheless, there are still countries, especially in emerging
economies, which do not actively promote the diffusion of renewables and, further, which maintain market barriers to their
dissemination, despite their countries' potential for use of renewables. In Latin American countries for example, the potential of
renewables including PV has been recognized at both utility and
residential scale, highlighting their benecial irradiance availability
and nancial competitiveness as compared with other technologies
[2,27,28]. Mexico, Brazil, Argentina, Chile, Peru, Colombia,
Guatemala, Ecuador and Venezuela were included in the European
Photovoltaic Industry Association (EPIA) PV opportunity mapping
of the Sunbelt, according to their solar resource and investment
attractiveness [28]. However PV installations remain low in most of
these countries, even though Mexico, Peru, Brazil and Chile stand

819

out because of their current installations and expected progress

[28,29].
An important reason for the low PV deployment in the Latin
American and Caribbean countries is, according to Jacobs et al. [29],
bad policy design that is not appropriate for the specic country
conditions or are non-attractive to investors. In this sense, Shirley
and Kammen [30] explain the importance of favourable policy,
designed according to the particulars of the countries concerned.
Some of the regulatory framework for renewables in Latin American countries includes Renewable Portfolio Standard or RPS (for
example, Peru), targets (such as in Argentina, Brazil, Colombia,
Nicaragua and Uruguay) and a combination of RPS, targets and a
renewable energy law (as in Chile and Mexico) [2]. In particular,
though Colombia promulgated law 1715 in May 2014 [16], it has not
been yet fully implemented.
This paper identies some of the effects of the penetration of
PVs on electricity markets, where there are favourable conditions
but no clear incentives. The focus here is on the number of
households that may adopt PV systems and on the load duration
curves as well as on the electricity tariff from the grid. These effects
are assessed under two scenarios: PV system with batteries as a
storage mechanism, and PV systems without storage but with the
support of electricity from the grid, aiming in both scenarios to
identifying changes in electricity demand patterns.
To analyse diffusion problems, there are two main alternative
levels of analysis reported in the literature: the micro and the
macro level [31]. At the micro level of analysis, the different authors
consider agent based, cellular automata and percolation modelling,
where the unit of analysis is a single individual considering his/her
perception of relative advantages and disadvantages of technology
adoption, and the inuence exerted by neighbours [31].
At the macro level, the unit of analysis is the aggregated social
system or community, and the focus is on market size and the
adoption time [31]. This perspective includes methodologies such
as system dynamics, which consider the combined complex
behaviour of a set of individuals characterized by their collective
behaviour, and represented by accumulated variables [32].
This paper uses the macro-level perspective, with the support of
a system dynamics model of several well-known system characteristics, including a) system complexity, b) variable accumulation,
c) feedbacks, and d) delays [33e35]. Furthermore, SD has been
largely used in policy assessment in related energy applications
[11,36e41], including photovoltaics policy analysis [42].
3. The Colombian electricity market
The Colombian power market is highly liberalized (open to
competition) through the electricity supply chain, except for the
household sector, which is regulated via a tariff mechanism and
supplied by the local utility [43]. As previously noted, the residential sector in Colombia consumes about 40% of the total electricity generated, and about 70% of residences are houses, rather
than ats in apartment blocks. Further, in Colombia, about 96% of
the population has regular access to electricity [44,45].
Average electricity consumption of households in Colombia
reaches 153.96 kWh/month (calculation with data from
Refs. [46,47]). Fig. 1 shows the average hourly load curve prole for
an average day in the Colombian Electricity Market, and Fig. 2
shows the specic hourly load curve prole for the residential
sector in Colombia.
The electricity tariff in the household sector for the main cities
was between USD 0.124 and USD 0.139 in 2014 (using data from
Ref. [17] and an exchange rate of 3000 COP/USD). This tariff is
calculated as the sum of: the generation price and the costs of
trading, losses, restrictions, and distribution and transmission

820

M. Jimenez et al. / Energy 111 (2016) 818e829

+
Incentive to invest
in Installed
Capacity

Price
Cost of Trading
Cost of Losses
Cost of Restrictions

+
+
+ Tariff
+
+
+

B1

R3
Cost of
Distribution
R4

+
Cost

Fig. 1. Daily load curve for the colombian electricity market.

Source: Authors, with data from Ref. [48].

Perceived
Benefit

Demand on
the System

Installed
Capacity

B3
-

Cost of
Transmission

+
R1

Households with PV
Microgeneration System
+

R2
Experience in PV
Panel Installation
+

Margin
+
-

B2

+
Installation of PV
Panels

Population

B4

Potential Households+of PV
Microgeneration System

Fig. 3. Dynamic hypothesis.

Source: Authors

Fig. 2. Daily load curve for the residential sector.

Source: Authors, with data from Refs. [48,49].

costs, where the distribution and transmission costs depend on the

total demand [50].
As solar PV diffuses, particularly into the household sector, this
load curve will be transformed, with unknown consequences for
the generation, transmission and distribution industries. Although
this has been assessed elsewhere, the likely prole of demand
needs to be better understood in order to assess the particular
consequences that this may entail for the value chain of the electricity industry in the case where there is no policy in place. This
paper makes progress investigating the potential benets of policy
provision. A model was constructed to provide initial insights into
the possible effects of PV penetration in the electricity industry. The
model is discussed in some detail in the following section.
4. Model description
This section discusses the model that was developed to support
the analysis, the main relationships and equations involved, and
justications for the assumptions as well as for the scenarios that
have been considered.
4.1. Dynamic hypothesis
Fig. 3 shows the proposed dynamic hypothesis that, through
eight loops, may explain the system behaviour and, in particular,
determine the transformation of the duration load curve of the
system. Broadly, this hypothesis indicates that the number of
households that adopt rooftop PV systems increases or decreases
according to the perceived benet, which depends on the

relationship between the cost of the PV system and the grid tariff. It
also explains that the number of households with PV systems affects system demand and consequently the grid tariff.
Loops B1 and B2 explain the price formation in electricity
markets as given by Dyner [51]; Price is affected negatively by the
system margin (Margin is the difference between Installed Capacity
and the Demand on the System); it means that when Margin increases Price decreases because low demand can be supplied using
cheaper technologies. System demand is negatively affected by
households that acquire microgeneration (as part of their demand
is generated by their own PV systems) and by electricity tariff.
Finally, Installed Capacity increases when there are incentives to
invest in installed capacity; and these incentives are directly linked
to Price: when the price is higher, there is more incentive to invest.
Loop R1 explains the feedback between households with a PV
system and the perceived benet, showing the positive causal effect between them. Also, loop B4 explains that Potential Households increases households that adopt PV systems, but these in turn
cause decreases in the variable Potential Households.
Households with a PV Microgeneration System affect positively
Installation of PV Panels which, in turn, increase the Experience in
PV Panel Installation (because of the learning curves); this experience will decrease the cost of microgeneration as we explain in
Loop R2. An increase in the cost of microgeneration causes a
decrease in the Perceived Benet of microgeneration acquisition,
which again affects Households with PV Microgeneration System.
Loops B3, R3 and R4 show how distribution and transmission
costs depend on the system's demand, where price, trading costs,
losses and restrictions costs all affect the grid tariff. As Tariff increases there is a perception of the benet of microgeneration,
which increases the number of households that micro-generate
and decreases the system demand, which has an effect on system
margin; Tariff is affected by transmission and distribution costs, on
top of other related costs, as indicated in Fig. 3.

4.2. Stock and ow diagram

Model simulation in this paper is undertaken by the system
dynamics approach, as discussed earlier. This methodology facilitates the comprehension of complex systems through a dynamic

M. Jimenez et al. / Energy 111 (2016) 818e829

hypothesis, as discussed in Section 4.1. As an aid to better understanding, this section describes the main model components using
stock and ow diagrams, and displays the main mathematical
equations.
Fig. 4 shows the population dynamics and explains that the
household stock in the system depends on the population growth
dynamics and the number of persons per household. The Population stock increases according to the Population Growth Rate, and
the Total Households stock increases with New Households, which
depends on the population growth and average Persons per
Household.
Fig. 5 describes the stocks and ows related to the demand
dynamics, which, as in the explanation of the dynamic hypothesis,
depends on the total microgeneration in the system and the
average electricity requirements of households. Residential Demand on the System is the difference between the total households'
electricity requirement and the total microgeneration produced by
the adopter households.
Fig. 6 shows the diffusion process using the Bass model. It indicates that households that have access to information become
potential adopters of PV systems, and that the decision to adopt the
technology depends on the relationship between the grid tariff and
the PV system cost, as explained in the dynamic hypothesis e the
adoption is modelled according to a logit model.
Finally, Fig. 7 indicates how tariff is determined by generation
price and the costs of restrictions, losses, trading, and transmission
and distribution costs. Transmission costs depend on the total
market demand, and distribution costs to households depend on
residential demand.

4.3. Model equations and assumptions

Table 1 presents the main model equations, their units and a
brief description of the equations.
The model boundary, considerations and assumptions, include:

Adoption of PV systems by the residential sector only.

All other electricity demand sectors, different from households,
increase according to estimates from Ref. [44].

Installed capacity (of PV systems) meets the average electricity
requirements of households (using systems of 1.5 kW). This
assumption implies that there could be times when electricity
generated by the PV system exceeds or falls short of household
requirements. Note that the current regulation does not allow
sales of PV surpluses to the grid; however, electricity decits
from the PVs are either met by electricity from the grid (when
connected) or using batteries (if not connected).

Fig. 4. Stock and ow diagram. Population Growth section.

Source: Authors

821

Consumption habits of households remain constant during the

simulation time.

The PV system efciency is the average of the PV system efciency during its lifetime.

Finally, the paper assumes that the national PV learning-curves
have the same behaviour as the international ones.
This paper analyses two scenarios: Storage Scenario and NoStorage Scenario. The rst one considers households that adopt a
PV system that includes batteries for electricity storage, for
whenever there is no sunshine. The second one considers a
standalone PV system (with no storage backup), but with grid
support, which means that households with microgeneration can
always get their energy shortfall from the grid. The simulation
timeframe is 14 years, from 2014 to 2028.
5. Results
Figs. 8e13 show output from simulation runs. Figs. 8 and 9 show
the diffusion process of PV systems with battery backup. Figs. 10
and 11 show how potential households become photovoltaic
adopters without storage capacity. Figs. 12 and 13 show the relationship between adopter households and total households in the
system, for both scenarios.
The potential household level is higher when storage is
considered than when no storage capacity is in place (Figs. 8 and
10). Although PV with storage capacity is more expensive than
electricity tariff during part of the period, the higher cost with
storage becomes relatively minor as learning curves reduce technology costs over time.
Results show the typical s-shaped diffusion behaviour (Figs. 9
and 11), peaking at about 2027, and reaching approximately six
million household adopters in both scenarios. These results show
that there is no signicant difference between the numbers of
adopters in the two scenarios (around one thousand households),
despite the initial differences in system prices. This opens the door
to other kinds of analysis concerning the importance of prices in the
adoption decision for this type of technology.
Finally, results in Figs. 12 and 13 show that at the end of the
simulation time the rate of adopter households compared to total
households in the system is similar in both scenarios, according to
results shown in Figs. 9 and 11.
System effects of the microgeneration diffusion process are
explained next, for each scenario.
5.1. Storage scenario
In this scenario, the PV system includes storage capacity. This

822

M. Jimenez et al. / Energy 111 (2016) 818e829

Fig. 5. Stock and ow diagram. Demand section.

Source: Authors

Fig. 6. Stock and ow diagram. Diffusion section.

Source: Authors

Fig. 7. Stock and ow diagram. Grid tariff section.

Source: Authors.

means that throughout a regular day, during periods of sunshine,

households can save any excess of generated electricity by using
batteries. The stored electricity can be used during the time when
there is no sunshine, which includes night time. That means that
the average household under this scenario does not use electricity
from the grid; however, given the present Colombian regulation,

households cannot sell their surpluses to the grid.

Fig. 14 shows simulations of expected increases in electricity
demand, over time, under a scenario where there is no adoption of
PV systems, which results from increases of new households over
the following years. Figs. 15 and 16 show the consequences on the
daily load curve when households adopt PV systems with storage in
place.
Under these conditions, when PV systems are available to
households, the residential duration load curve decreases in parallel because of the use of electricity from microgeneration sources
during the whole day (Fig. 16). However the whole market demand
for electricity continues growing over time, although more slowly
than in the base case scenario, when no PV adoption takes place
(Fig. 15). This result is explained by the participation of the
household sector in the system demand, which accounts for about
40% of demand, assuming no PV adoption by the remaining sectors
(i.e. industrial, commercial and government) e taking into account
the fact that any surpluses of electricity generated by households
cannot supply other sectors, because current regulation forbids
households selling surpluses to the market. If such sales were
allowed, the load curve prole of the whole market would be
different, as the sale of excess generation by households supplying

M. Jimenez et al. / Energy 111 (2016) 818e829

823

Table 1
Model equations.
Equations and comments

Units

Potential Households:

Households

Rt
Potential Households 0 Perceived Benefit Bass  Cc Adoptiondt (1)
This stock considers the number of households that perceived the benets of PV microgeneration, excluding adopter households.
Adopter Households:
Rt
Adopter Households 0 Cc Adoptiondt (2)
This stock considers the number of household adopters in the system.
CC_Adoption:
g

Cc Adoption Potential Households*A:CostA:Cost

g A:Tariff g (3)

Households

Households/
month

The Coefcient of Adoption was modelled according to a Logit model that includes the cost of PV microgeneration, the grid Tariff and the parameter
gamma [52].
kWh/month
Residential demand on the System:
R:Demand H:Requirement per Household*Total Households  Total Microgeneration (4)
It is the total kWh/month that the households demand of the system. It is calculated as the total electricity requirement excepting the electricity that is
covered by the total microgeneration of the system.

Fig. 8. Potential households e storage scenario.

Source: Authors/simulation runs.

Fig. 10. Potential households e no-storage scenario.

Source: Authors/simulation runs

Fig. 9. Adopters' households e storage scenario.

Source: Authors/simulation runs.

Fig. 11. Adopters' households e no-storage scenario.

Source: Authors/simulation runs.

to other sectors would reduce demand from the grid's main

(commercial) generators.
Fig. 17 shows the consequences of this scenario on the hourly
electricity tariff. As previously discussed, the tariff depends on
generation price, and the costs of trading, losses, restrictions, distribution and transmission. Also note that as transmission and

distribution costs (Figs. 19 and 20) depend on system demand, this

implies that as demand drops so grid tariff increases, because costs
have to be paid by fewer households. The opposite effect occurs for
the generation component of the tariff (generation price), as this is
the lowest price at which system demand can be satised. When
demand is low, generation cost falls, because a low demand could

824

M. Jimenez et al. / Energy 111 (2016) 818e829

Fig. 15. Hourly load curve, Colombian Market e Storage Scenario. (Arrow indicates
direction of time).
Source: Authors/simulation runs.

Fig. 12. Relationship between adopters' households vs total households e no-storage

scenario.
Source: Authors/simulation runs.

Fig. 16. Hourly load curve, Residential Sector e Storage Scenario. (Arrow indicates
direction of time).
Source: Authors/simulation runs.

Fig. 13. Relationship between adopters' households vs total households e storage

scenario.
Source: Authors/simulation runs.

be completely satised by generation from a low cost technology

(like hydro) and does not require the operation of more expensive
technologies such as those that are thermoelectric-based (Fig. 18).
As a result of this, grid tariff changes over time according to the
balance of the different effects on its components, as visible in
Fig. 17.
There is a signicant effect of the penetration of solar PV into
households on the hourly electricity tariff, because of the intermittency of solar power during the day. Hourly tariffs are highest at
around seven o'clock at night and lowest at two or three o'clock in
the morning, at the beginning of the simulation, when PV penetration into households is only just starting. The hourly electricity
tariffs gradually become lower as the market for solar PV approaches saturation and the rate of penetration diminishes. When
installation of solar PV has reached its maximum (i.e. when penetration is highest) then the duration load curve converges to a
constant (0.12USD) as shown in Fig. 18. However, changes in the
residential system-demand lead to a distribution cost increase over
time, until it becomes large enough that the tariff must increase
(Fig. 20). Transmission cost depends on total demand (not only on
the demand of households); while households leave the grid (up to
a limit) demand from other sectors keeps increasing, resulting in
oscillations, as visible in Fig. 19.

5.2. No-storage scenario

Fig. 14. Hourly load curve without residential PV adoption. (Arrow indicates direction
of time).
Source: Authors/simulation runs.

In this scenario, the penetration of PV systems does not include

batteries for electricity support, which means that households use

M. Jimenez et al. / Energy 111 (2016) 818e829

825

Fig. 20. Distribution cost e storage scenario.

Source: Authors/simulation runs.
Fig. 17. Hourly grid tariff e storage scenario.
Source: Authors/simulation runs.

Fig. 18. Hourly generation price e storage scenario.

Source: Authors/simulation runs.

grid decreases in the residential sector (Fig. 21) and in the whole
market (Figs. 22 and 23, respectively). For the residential sector, this
may even become negative under particular circumstances, which
means that there is excess capacity in the system. However, during
dark periods of the day, the system demand increases each year,
because the total number of households keeps increasing over time.
Note that electricity required during dark periods is supplied from
the grid.
The results for the entire market show that if regulation allowed
households' excess capacity to be used to supply other demand
sectors, during periods of sunshine, the grid-demand would be
reduced (Fig. 22). The consequences of this behaviour on the grid
tariff are presented in Fig. 24.
Fig. 24 shows the hourly evolution of the grid tariff e this, the
Colombian swan gure. Note that there are increases up to 250%
higher than for the storage scenario, over the simulation period.
These increases result from higher adoption rates of PVs, leading to
large increases in distribution costs, which depend on the residential demand (the larger the demand from the grid, the lower the
tariff). The behaviour of the components of the grid tariff is shown
in Figs. 25e27.
Note that the distribution cost increases, during the simulation
time, up to a limit that is required in order to prevent system
collapse (Fig. 27). Transmission cost also increases up to a limit
(Fig. 26). However, generation prices decrease because of the
reduction in the system demand requiring the cheapest units for
operation, but they increase at the end of the period because of the
distribution restrictions after around 2022 (Fig. 25).
These results shows that if no regulation is implemented in the
system, adoption with no storage would cause the system to crash,
as tariffs reach 0.37 USD/kWh, which is unsustainable. Moreover,
these results show that under these conditions the market collapses around 2024.

6. Validation and sensitivity analysis

Fig. 19. Transmission cost e Storage Scenario.
Source: Authors/simulation runs.

electricity from the grid when there is no sunshine. Consequently,

the system load curve is reduced only during times when there is
sunshine. This produces the well-known duck graph of the system e in this case the Colombian duck load curve, as shown in
Figs. 21 and 22.
Figs. 21e23 show that when the sun shines, demand from the

This section reports structural and behavioural validation processes undertaken in order to assess model consistency and
robustness, aimed at gaining condence in the simulations [33,53].
Moreover, this section also assesses some of the most relevant assumptions and initial model conditions, as well as sensitivity
analysis of results. For validating model structure, two different
tests were conducted. First, for boundary adequacy, testing examines whether endogenous and exogenous variables are consistent
with the purpose of the model. In this case, endogenous variables
such as Households with and without Knowledge, Adopter

826

M. Jimenez et al. / Energy 111 (2016) 818e829

Fig. 21. Hourly load curve. Residential sector e no-storage scenario.

Source: Authors/simulation runs.
Fig. 24. Hourly grid tariff. The Colombian swan e no-storage scenario.
Source: Authors/simulation runs

Fig. 22. Hourly load curve for Colombian market using the surpluses of residential
generation e no-storage scenario.
Source: Authors/simulation runs.

Fig. 25. Hourly generation price e No-Storage scenario.

Source: Authors/simulation runs.

Fig. 23. Hourly load curve for Colombian market losing surpluses of residential generation e no-storage scenario.
Source: Authors/simulation runs.

Households, Total Microgeneration in the System, Residential Demand on the System, Tariff, generation Price, and Distribution and
Transmission costs evolve according to their feedbacks, which is
consistent with the purpose of the model. Additionally, variables
such as Cost of electricity Losses, Cost of Restrictions, Trading Costs,
Hourly and Monthly electricity Requirement per Household, among
others, are exogenous given the purpose of the model. Second,
under the extreme-condition test, model runs showed that no PV
adoption took place: a) when grid tariff was free or negligible, and
b) when population dynamics was non-existent.

Fig. 26. Transmission cost e No-Storage Scenario.

Source: Authors/simulation runs.

These tests indicate structural consistency. For behavioural

validation, this paper undertakes sensitivity analysis. First, because
model t to data was impossible for the Colombian case as historical data is not available for error analysis tests, given that policy for
the promotion of PVs is not in place yet, thus data is not available
for statistical analysis. Second, as the paper is forward-looking,
analysis focuses on future behaviours and their consequences for

M. Jimenez et al. / Energy 111 (2016) 818e829

Fig. 27. Distribution cost e No-Storage Scenario.

Source: Authors/simulation runs

system transformation. Thus, for robustness and soundness purposes, this paper undertakes sensitivity analysis of parameters, as
reported below.
The rst evaluated parameter is the innovator's parameter (p),
which has been modied by 50% upwards and downwards from its
initial value. Results are presented in Fig. 28.
Results indicate that the value of the innovator's factor, p, is
important for dening the time when adoption begins. When the
innovator's factor is larger the diffusion process starts earlier than
when this is lower; this is reasonable, considering that this factor
describes innovator willingness and that information is available to
him/her, increasing the adoption rate by households. It is important
to note that regardless of the value of p at the end of the simulation
time, the number of adopters is the same, which gives robustness to
the results presented in Section 5 and reliability to the model
ndings.
The second condition evaluated is the sensitivity to the price
conditions. In this case, a comparison is made with the adoption
level when the PV system cost is changed. To make this comparison
it is important to discuss the importance of the technology price for
adoption in a place with the Colombian conditions. The literature
reports that nancial issues rather than incentives become barriers
to the adoption of renewables [10]. Moreover, Claudy, Michelsen,
O'Driscoll, and Mullen [54] highlight the importance of technology
information over the nancial issues, and explain that the technology knowledge and awareness precede the product characteristics' evaluation. Other authors explain that there are other factors

Fig. 28. Adopter households' sensitivity to value of innovator's parameter p.

Source: Authors/simulation runs.

827

that are more important than the price difference [55], and authors
like Islam [56] highlight the importance of education and information over the adoption price difference. Most important to note
is that, in this case, grid parity was attained very early in the
simulation period; Fig. 29 shows simulation results for the sensitivity analysis of the costs of photovoltaic systems that were revised
50% upwards and downwards, respectively.
This gure shows that even for extreme cases, differences in the
rate of adoption of PVs remain negligible when grid parity has been
attained (as is the case in many Colombian sites [15]) and PV prices
are highly competitive compared with the grid tariff, as the
magnitude of the difference between the two is very large.
Finally, what may happen when grid parity is only reached
around 2022, half way through the simulation period, i.e. in places
where grid parity has not been attained yet? Three different alternatives are presented in Fig. 30, where the main conclusion is
that grid parity only delays adoption rates during the initial period,
but adoption levels converge to the same value at the end of the
simulation.
7. Conclusions
Favourable conditions for PV systems e sunshine availability
and low PV microgeneration cost [15] e will contribute to rapid
technology diffusion in the Colombian residential sector, as illustrated in this paper. This will, however, have a great effect on grid
electricity costs to non-PV users and will pose challenges to system
operators, as they will confront complicated load duration curves.
The effect of the diffusion of PV systems on tariffs and load
duration curves will depend on the chosen scenario. On the one
hand, under the storage scenario, the system exhibits gradual decreases in residential electricity demand and smooth variations in
electricity demand for the entire electricity market, which is reected in the load duration curve; furthermore, there are smooth
decreases in grid tariffs when PVs initially start penetrating the
market, but increases at the end of the simulation period as distribution cost exceeds generation. On the other hand, for the NoStorage Scenario, the entire system confronts dramatic changes,
including extreme increases in grid tariffs and problems in the
system balances, as can be appreciated from the load duration
curve.
Comparing both scenarios, we observe rst that the adoption
rates are similar in both scenarios despite the differences in PV
system costs, which can suggest that nance might not be the main

Fig. 29. Adopter households' sensitivity to PV system cost.

Source: Authors/simulation runs

828

M. Jimenez et al. / Energy 111 (2016) 818e829

Table A1
Abbreviations.

Fig. 30. Sensitivity analysis when grid parity is achieved half way through the simulation period.
Source: Authors/simulation runs

reason for adopting this type of technology in the Colombian case.

Furthermore, this paper establishes that both scenarios are
complicated in terms of system stability, and that the inclusion of
batteries only delays the problem of achieving system stability.
However the No-Storage Scenario can be described as the worstcase scenario, because it shows more extreme results in the
simulation time, with the important peak increases in the load
duration curve, and the need for a signicant amount of electricity
from the grid in periods without sunshine, demanding a high
backup capacity from the grid.
Results indicate that: a) PV system cost is not necessarily an
important element in the diffusion of PV systems when grid parity
is attained, b) the design and implementation of policies is an
imminent obligation for policymakers in the market because the
lack of policy may cause catastrophic effects on the market, risking
not only market stability but also its sustainability.
Although policy is not the focus of this paper, some insights
might be gained. First, this paper shows that Colombia has a great
potential for the deployment of PVs in the household sector and in
this sense could follow or overtake the diffusion levels of other
Latin American countries such as Mexico, Brazil or Chile, or even
leading countries elsewhere. Second, policymakers ought to be
aware that it may not be wise to just wait until PV systems attain
grid parity before initiating regulation, as by then the PV systems
may well have already had effects on the grid that need attention e
the system operation by then could be completely different as
might be deduced from the resulting load duration curve. This calls
for further research as this paper emphasizes that having no policy
at all is unwise for electricity markets in the developing world, and
likewise in the developed world. Though the attention has been on
the Colombian case, much of the benet from the insights can be
extended to other places e particularly, to that part of the world
where grid parity of PVs on rooftops is being achieved.

Acknowledgments
The authors are grateful to EPM S.A. E.S.P (CIIEN N
29990835044), COLCIENCIAS (Grant 645, 2014) and Universidad
Nacional de Colombia (Grants CF 161-163 and CF 179-177) for
providing the necessary nancial support for this paper. The authors are grateful to Dr Paul Ellis for proofreading.

Annex
Table A1 shows terms and abbreviations used in the paper.

Abbreviation

Explanation

COP
EPIA
FIT
GW
IEA
kWh
p
PV
q
RPS
SD
TWh
USD

Colombian Pesos
European Photovoltaic Industry Association
Feed-in Tariff
Giga watt
International Energy Agency
Kilowatt-hour
Innovation parameter
Photovoltaics
Imitation parameter
Renewable Portfolio Standard
System dynamics
Terawatt-hour
United States Dollar

References
[1] Technology Roadmap IEA. Solar photovoltaic energy. 2014th ed. Paris: International Energy Agency; 2014.
[2] IEA. Medium-term renewable energy market report 2014. 2014th ed. Paris:
International Energy Agency; 2014. http://dx.doi.org/10.1787/renewmar2014-en.
[3] Solangi K, Islam M, Saidur R. A review on global solar energy policy. Renew
Sustain
Energy
Rev
2011;15:2149e63.
http://dx.doi.org/10.1016/
j.rser.2011.01.007.
[4] Balcombe P, Rigby D, Azapagic A. Motivations and barriers associated with
adopting microgeneration energy technologies in the UK. Renew Sustain
Energy Rev 2013;22:655e66. http://dx.doi.org/10.1016/j.rser.2013.02.012.
[5] Teske S, Pregger T, Simon S, Naegler T, Graus W, Lins C. Energy [R]evolution
2010-a sustainable world energy outlook. Energy Efc 2011;4:409e33. http://
dx.doi.org/10.1007/s12053-010-9098-y.
[6] Hurtado Munoz LA, Huijben JCCM, Verhees B, Verbong GPJ. The power of grid
parity: a discursive approach. Technol Forecast Soc Change 2014;87:179e90.
http://dx.doi.org/10.1016/j.techfore.2013.12.012.
[7] Chowdhury S, Sumita U, Islam A, Bedja I. Importance of policy for energy
system transformation: diffusion of PV technology in Japan and Germany.
Energy
Policy
2014;68:285e93.
http://dx.doi.org/10.1016/
j.enpol.2014.01.023.
[8] Jacobsson S, Lauber V. The politics and policy of energy system transformationdexplaining the German diffusion of renewable energy technology.
Energy
Policy
2006;34:256e76.
http://dx.doi.org/10.1016/
j.enpol.2004.08.029.

[9] del Ro Gonzalez P. Ten years of renewable electricity policies in Spain: an
analysis of successive feed-in tariff reforms. Energy Policy 2008;36:2917e29.
http://dx.doi.org/10.1016/j.enpol.2008.03.025.
[10] Balcombe P, Rigby D, Azapagic A. Investigating the importance of motivations
and barriers related to microgeneration uptake in the UK. Appl Energy
2014;130:403e18. http://dx.doi.org/10.1016/j.apenergy.2014.05.047.
[11] Franco CJ, Castaneda M, Dyner I. Simulating the new British electricity-market
reform. Eur J Oper Res 2015. http://dx.doi.org/10.1016/j.ejor.2015.02.040.
n JC, Rajoy Brey M. Suspensio
n de los procedimientos de pre[12] de Borbo
n de retribucio
n en las nuevas instalaciones de produccio
 n de
asignacio
ctrica en re
gimen especial. 2012. p. 8068e72.
energa ele
[13] DECC. Ensuring value for money and maintaining investment in renewable
energy. News Story; 2014. https://www.gov.uk/government/news/ensuringvalue-for-money-and-maintaining-investment-in-renewable-energy.
[14] DECC. UK solar PV strategy Part 2: delivering a brighter future. 2014. p. 58.
nez M, Cadavid L, Franco C. Scenarios of photovoltaic grid parity in
[15] Jime
Colombia.
Dyna
2014;188:237e45.
http://dx.doi.org/10.15446/
dyna.v81n188.42165.
[16] Congreso de la Repblica de Colombia. Ley 1715. Por la cual se regula la
n de las energas renovables no convencionales al Sistema Enerintegracio
tico Nacional. Colombia. 2014.
ge
n de Servicios Pblicos (SUI). Consolidado
[17] SUI. Sistema nico de Informacio
Energa Por Empres Dep Y Munic Consum Promedio. 2014. http://reportes.sui.
gov.co/fabricaReportes/frameSet.jsp?idreporteele_com_096.
[18] Boletn DANE. Censo general 2005. Bogot
a: Perl Colombia; 2010.
[19] Balcombe P, Rigby D, Azapagic A. Environmental impacts of microgeneration:
integrating solar PV, Stirling engine CHP and battery storage. Appl Energy
2015;139:245e59. http://dx.doi.org/10.1016/j.apenergy.2014.11.034.
[20] der Veen RAC, De Vries LJ. The impact of microgeneration upon the Dutch
balancing market. Energy Policy 2009;37:2788e97. http://dx.doi.org/10.1016/
j.enpol.2009.03.015.
n J, Fras Marn P, Hakvoort R. The economic effect of
[21] Eid C, Reneses Guille
electricity net-metering with solar PV: consequences for network cost recovery, cross subsidies and policy objectives. Energy Policy 2014;75:244e54.
http://dx.doi.org/10.1016/j.enpol.2014.09.011.
n J. Improved photovoltaic self-consumption with appliance scheduling
[22] Wide

M. Jimenez et al. / Energy 111 (2016) 818e829

[23]

[24]

[25]

[26]

[27]

[28]
[29]

[30]

[31]

[32]

[33]
[34]

[35]

[36]

[37]
[38]

in 200 single-family buildings. Appl Energy 2014;126:199e212. http://

dx.doi.org/10.1016/j.apenergy.2014.04.008.
Keirstead J. Behavioural responses to photovoltaic systems in the UK domestic
sector.
Energy
Policy
2007;35:4128e41.
http://dx.doi.org/10.1016/
j.enpol.2007.02.019.
Zuluaga MM, Dyner I. Incentives for renewable energy in reformed LatinAmerican electricity markets: the Colombian case. J Clean Prod 2007;15:
153e62. http://dx.doi.org/10.1016/j.jclepro.2005.12.014.
Duan H-B, Zhu L, Fan Y. A cross-country study on the relationship between
diffusion of wind and photovoltaic solar technology. Technol Forecast Soc
Change 2014;83:156e69. http://dx.doi.org/10.1016/j.techfore.2013.07.005.
Agterbosch S, Meertens RM, Vermeulen WJV. The relative importance of social
and institutional conditions in the planning of wind power projects. Renew
Sustain
Energy
Rev
2009;13:393e405.
http://dx.doi.org/10.1016/
j.rser.2007.10.010.
vez DL, Daz-Chavez R,
Alem
an-Nava GS, Casiano-Flores VH, C
ardenas-Cha
Scarlat N, Mahlknecht J, et al. Renewable energy research progress in Mexico:
a review. Renew Sustain Energy Rev 2014;32:140e53. http://dx.doi.org/
10.1016/j.rser.2014.01.004.
EPIA. Global market outlook for photovoltaics. 2014. Brussels, Belgium.
Jacobs D, Marzolf N, Paredes JR, Rickerson W, Flynn H, Becker-Birck C, et al.
Analysis of renewable energy incentives in the Latin America and Caribbean
region: the feed-in tariff case. Energy Policy 2013;60:601e10. http://
dx.doi.org/10.1016/j.enpol.2012.09.024.
Shirley R, Kammen D. Renewable energy sector development in the Caribbean: current trends and lessons from history. Energy Policy 2013;57:
244e52. http://dx.doi.org/10.1016/j.enpol.2013.01.049.
Laciana CE, Rovere SL, Podest
a GP. Exploring associations between micro-level
models of innovation diffusion and emerging macro-level adoption patterns.
Phys A Stat Mech Appl 2013;392:1873e84. http://dx.doi.org/10.1016/
j.physa.2012.12.023.
Davidsson P, Verhagen H. Types of simulation. In: Edmonds B, Meyer R, editors. Simulating soc. complex. Berlin, Heidelberg: Springer Berlin Heidelberg;
2013. p. 23e36. http://dx.doi.org/10.1007/978-3-540-93813-2.
Sterman JD. Business dynamics. Systems thinking and modeling for a complex
world. United States: McGraw-Hill Higher Education; 2000.
Dyner I, Franco CJ. Consumers' bounded rationality: the case of competitive
energy markets. Syst Res Behav Sci 2004;21:373e89. http://dx.doi.org/
10.1002/sres.644.
Maier FH. New product diffusion models in innovation managementda system dynamics perspective. Syst Dyn Rev 1998;14:285e308. http://dx.doi.org/
10.1002/(SICI)1099-1727(199824)14:4<285::AID-SDR153>3.0.CO;2-F.
Bunn DW, Larsen ER. Sensitivity of reserve margin to factors inuencing investment behaviour in the electricity market of England and Wales. Energy
Policy 1992;20:420e9. http://dx.doi.org/10.1016/0301-4215(92)90063-8.
Ford A. Using simulation for policy evaluation in the electric utility industry.
Simulation 1983;40:85e92.
Ochoa C, Dyner I, Franco CJ. Simulating power integration in Latin America to

[39]

[40]
[41]

[42]

[43]

[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]

[53]

[54]

[55]

[56]

829

assess challenges, opportunities, and threats. Energy Policy 2013;61:267e73.

http://dx.doi.org/10.1016/j.enpol.2013.07.029.
Larsen ER, Dyner I, Bedoya L, Franco CJ. Lessons from deregulation in
Colombia: successes, failures and the way ahead. Energy Policy 2004;32:
1767e80. http://dx.doi.org/10.1016/S0301-4215(03)00167-8.
Naill RF. A system dynamics model for national energy policy planning. Syst
Dyn Rev 1992;8:1e19. http://dx.doi.org/10.1002/sdr.4260080102.
Ford A, Vogstad K, Flynn H. Simulating price patterns for tradable green
certicates to promote electricity generation from wind. Energy Policy
2007;35:91e111. http://dx.doi.org/10.1016/j.enpol.2005.10.014.
Movilla S, Miguel LJ, Bl
azquez LF. A system dynamics approach for the
photovoltaic energy market in Spain. Energy Policy 2013;60:142e54. http://
dx.doi.org/10.1016/j.enpol.2013.04.072.
Congreso de la Repblica de Colombia. Ley 143 de 1994. Por la cual se estagimen para la generacio
n, interconexio
n, transmisio
n, distribucio
n
blece el re
 n de electricidad en el territorio nacional, se conceden unas
y comercializacio
autorizaciones y se dictan otras disposiciones en materia ener, 1994; 1994.
Colombia.
 n de Referencia Generacio
 n-Transmisio
 n 2014e2028.
UPME. Plan de Expansio
2015. p. 765.
tico nacional Colombia: ideario energe
tico 2015. 2015.
UPME. Plan energe
p. 184.
.
DANE. Encuesta Nacional de Calidad de vidae2014. 2014. Bogota
Xm. Demanda de Energa 2014. 2015. http://www.xm.com.co/Pages/
DemandaEnergia-2014.aspx (accessed 15.08.15).
 n inteligente. 2015. http://informacioninteligente10.xm.com.
Xm. Informacio
co/pages/default.aspx (accessed 01.01.15).
n de demanda de energa ele
ctrica y potencia. Revisio
n
UPME. Proyeccio
.
marzo de 2008. 2008. Bogota
n 097 de 2008. 2008. Colombia.
CREG. Resolucio
Dyner I. Energy modelling platforms for policy and strategy support. J Oper
Res Soc 2000;51:136e44.
Fox JT, Kim K il, Ryan SP, Bajari P. The random coefcients logit model is
identied.
J
Econom
2012;166:204e12.
http://dx.doi.org/10.1016/
j.jeconom.2011.09.002.
Barlas Y. Formal aspects of model validity and validation in system dynamics.
Syst Dyn Rev 1996;12:183e210. http://dx.doi.org/10.1002/(SICI)10991727(199623)12:3<183::AID-SDR103>3.0.CO;2e4.
Claudy MC, Michelsen C, O'Driscoll A, Mullen MR. Consumer awareness in the
adoption of microgeneration technologies: an empirical investigation in the
Republic of Ireland. Renew Sustain Energy Rev 2010;14:2154e60. http://
dx.doi.org/10.1016/j.rser.2010.03.028.
Karakaya E, Hidalgo A, Nuur C. Motivators for adoption of photovoltaic systems at grid parity: a case study from Southern Germany. Renew Sustain
Energy Rev 2015;43:1090e8. http://dx.doi.org/10.1016/j.rser.2014.11.077.
Islam T. Household level innovation diffusion model of photo-voltaic (PV)
solar cells from stated preference data. Energy Policy 2014;65:340e50. http://
dx.doi.org/10.1016/j.enpol.2013.10.004.